Error Reporting Settings Block for React

First the ignore patterns that are switched on: if any pattern is contained in the message, it is dropped and the verdict names the pattern and the reason. Then the grouping rules that are switched on, from the top: the first whose pattern the message contains files it under that rule's group, and the verdict says which rule it was by its place in the list. If nothing matches it would start a new group. Matching ignores case, and a star in a pattern stands for any run of characters, so "Loading chunk * failed" matches "Loading chunk 482 failed".

Yes: the tester reads the rules as they are on the page, so switching a rule off or moving one up changes the verdict at once, before anything is saved. A few moments after you stop typing, or after a change that alters the verdict, the status line says it too, so a screen reader hears the result once instead of on every key. Pressing Enter in the tester tries the message without saving.

Edit the constants at the top. GROUP_RULES and IGNORE_RULES list the rules (an id, the group or reason in name, a pattern and the week's count), ENVIRONMENTS the environments with the events they send in a day, and SAMPLES the buttons that fill the tester. SAVED holds the order and the switches the account has stored, and the sample rates. Counts are shown as typed; take them from your own stats.

All of it: no error is received or dropped, the counts are typed, and Save keeps the rules in this page and says so. To make it real, send the ordered rules and rates to your ingest service when saving, run its own grouping on the tester's message (or the same function on the server), and take the weekly counts from its stats.

Yes. The tester's field is named and describes itself with the verdict, the sample buttons are real buttons, each rule has a named switch, and Move up and Move down are real buttons named for the rule ("Move Slow upstream down"), with the first rule's Up button and the last rule's Down button staying focusable but saying there is nowhere to go. After a move, focus stays on the button you pressed. The tabs follow the arrow keys.

The tester's field and samples stack at full width and the verdict box reserves its room. The three tabs share the width, and all their panels are as tall as the tallest, so changing tabs never moves the foot, which stays pinned to the bottom of the screen. A long pattern wraps rather than scrolls.